Wrong CPU power readings after sleep state

[Patishi]

Hi,
Before I am getting into more details, I would like to say that I am not sure if this is a HWiNFO issue or my motherboard's fault.
I am running windows 11 24H2 (build 26100). My motherboard is Asrock Z790 Riptide Wifi, CPU is 14900K. Bios updated to version 11.01 which have the latest 12B microcode.
My bios is set to static 253W max PL1 + PL2 (Intel default). I am checking this with Stockfish chess engine analysis software which loads 100% percent of the CPU.

After a normal system start (either cold start or Restart) the CPU package power reading never exceeds 253W at full load, but after the computer awakes from sleep I noticed that the power reading can reach above 253W (260-270+).
So either there is a bug from the software's side (HWiNFO/ Windows) or my motherboard is not respecting the power limits after it wakes up from sleep.
I thought maybe I should bring up this issue, maybe others encountered a similar issue? I will also report this to Asrock.

Thx.

UPDATE: I discovered that also after a cold start I get the same issue.. so my mistake. But after a restart everything is ok. This immediately led me to suspect Windows' "Fast startup" feature from the power options. So I turned it off and now the readings are fine either way. So this is the root of the problem.
But I still wonder if this is only a software thing (wrong reading) or the motherboard is really exceeding the power limits to the CPU.

 

[Martin]

Power reading works using a differential method - there are 2 energy samples collected at different time points (polling cycles) and the difference between them divided by time delta is the power result.
A system sleep or reboot turns off (or resets) several blocks in the CPU so if one sample was taken before sleep and the other one after, the result can be wrong.
IMO there's nothing wrong about this, such results can be expected in such case.

 

[Patishi]

Power reading works using a differential method - there are 2 energy samples collected at different time points (polling cycles) and the difference between them divided by time delta is the power result.
A system sleep or reboot turns off (or resets) several blocks in the CPU so if one sample was taken before sleep and the other one after, the result can be wrong.
IMO there's nothing wrong about this, such results can be expected in such case.

I understand what you say (well..as well as I can LOL). So you think this is only a reading issue? and the actual package power doesn't exceed 253W?
And how come it doesn't happen with Fast startup off?
The reason I came up with this is because I know that reboot doesn't use the fast startup, only a full shutdown (Microsoft also indicates it in the description).
So when I first noticed this issue after my PC was waking up from sleep, I checked if it also happens when I fully shut down the machine.. and it did.
This is because sleep mode is similar to shutting down the machine and fast startup feature is being used.

Just to make clear, the 253W limit by the motherboard supposed to be a hard brick, means that 253.09 is the max I see in full load operation (more correctly fluctuates between 252 and 253).
But it never goes beyond 253.
And also important to mention is that the PL1 and PL2 reading is always a static 253W. It's the actual "CPU package power" number that does this.
I attached a screen image so you see how it looks like in normal state.

It's important to know if windows Fast startup is somehow doing something to the motherboard and it doesn't stays within the power limits. This is something that need to be checked by motherboard's manufacturers. Because it's definitely not supposed to happen.

 

[Martin]

Fast boot is a software feature, it should have no impact on hardware power monitoring. The "inaccuracy" of power reporting is due to transient changes to internal counters/timer.

 

[Patishi]

Fast boot is a software feature, it should have no impact on hardware power monitoring. The "inaccuracy" of power reporting is due to transient changes to internal counters/timer.

But if Fast startup is off I don't get these "inaccuracies". So it must be related somehow.